16th World Conference on Earthquake, 16WCEE 2017

Santiago Chile, January 9th to 13th 2017

Paper N 0900 (Abstract ID)
Registration Code: S-W1463067441

TORSIONAL RESPONSE OF HORIZONTALLY CURVED BRIDGES

SUBJECTED TO EARTHQUAKE-INDUCED POUNDING

M. Amjadian (1), A. K. Agrawal (2)

(1)
PhD Candidate, The City College of the City University of New York, mamjadian@ccny.cuny.edu.
(2)
Professor, The City College of the City University of New York, agrawal@ccny.cuny.edu.

Abstract
Horizontally curved bridges have been observed to be highly vulnerable to severe seismic damage, such as deck unseating,
because of the strong tendency of their decks to in-plane rotation. The torsional demand of these bridges may dramatically
increases when they are subjected to earthquake-induced pounding, as observed in Baihua and Huilan bridges during 2008
Wenchuan earthquake in China. This paper is focused on studying the influence of seismic pounding, deck-abutment
collision, on the torsional response of horizontally curved bridges in strong earthquakes. For this purposes, a three-degree-
of-freedom nonlinear model has been developed to capture key dynamic parameters affecting the seismic response of these
types of bridges. In the model, the radial and azimuthal shear-displacement relations of the columns and their bilateral
interactions are modeled by a coupled-biaxial bilinear hysteresis spring, but their torsional moment-rotation relations are
molded by a linear spring; and the normal and tangential impact forces at the contact nodes between the deck and the
abutments are modeled by Jankowski contact model and Karnopp friction model, respectively. Then, the nonlinear bridge
model is utilized to perform a detailed parametric analysis on the torsional response of a prototype curved bridge based on
variations of these key parameters of the model: (i) subtended angle of the deck, ; (ii) size of the gap between the deck and
the abutments, g ; (iii) normalized static radial stiffness eccentricity, sr . The results obtained from this numerical analysis
indicate that the seismic pounding has a noticeable effect on the in-plane rotation of the curved bridge deck particularly
when the distance between the stiffness center and the curvature center of the deck is increased.
Keywords: Horizontally Curved Bridge; Seismic Pounding; Torsional Response; Non-Linear Model; Parametric Analysis.
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

1. Introduction
In the recent decades, horizontally curved bridges have been widely used in urban transportation networks
around the world. The curved geometries of these bridges enable them to be suitably located in complex grade-
separated intersections and interchanges where there is a strong emphasis on aesthetic and space-compatible
structural designs at the same time [1,2]. However, due to their irregular geometries and asymmetric stiffness
and non-uniform mass distributions, horizontally curved bridges, similar to other types of irregular bridges such
as skew bridges [35], are more susceptible to seismic failure than straight bridges [68]. It is also believed that
multi-support irregular bridges, such as skew and curved bridges, are more susceptible to the rotational
components of strong ground motions [9,10]. The unique structural features, inherent in horizontally curved
bridges, have significant impacts on the amplification of the torsional response of the decks of these bridges
particularly when they collide with the abutments during strong earthquakes (i.e. earthquake-induced pounding).
The impact forces resulted from the collision will likely cause large radial and azimuthal displacements in the
bearings at the abutments, and consequently, unseating of the decks from the abutments [1]. For example, it has
been reported that seismic pounding was an effective factor in structural failures of Baihua and Huilan bridges
during 2008 Wenchuan earthquake in China with a high moment magnitude of M W =7.9 [1113].
A large number of horizontally curved bridges either collapsed or suffered serious damage in the great
1971 San Fernando earthquake in California [7]; and since then, the vulnerability of these bridges to seismic
damage has been a subject of interest to many researchers. However, there have been few studies on the
vulnerability of horizontally curved bridges to earthquake-induced pounding. Williams and Godden [14]
conducted an experimental study on the seismic response of a small scale model of a curved RC bridge that
suffered extensive damage during the 1971 San Fernando earthquake. The authors considered the effects of
sliding and seismic pounding at the expansion joints and ductility in the columns. Ijima et al. [15] used analytical
and experimental models to study the influence of seismic pounding on the collapse of the decks of skew and
curved bridges. The study was inspired by observation of the deck unseating in these bridges because of seismic
pounding in the 1995 Kobe earthquake in Japan. Wieser et al. [16] performed an experimental test on a 2/5 scale
model to investigate the effects of deck-abutment collision on the overall seismic performance of horizontally
curved bridges. Ruiz Julian et al. [17] studied the efficiency of cable restrainers in the protection of curved steel
viaducts from deck unseating caused by seismic pounding. Recently, Amjadian and Agrawal [1] have carried out
an extensive parametric study on the influence of seismic pounding on the rigid-body motion of horizontally
symmetric curved bridges.
This paper, using a three-degree-of-freedom nonlinear model, studies the sensitivity of the torsional
response of horizontally one-way-asymmetric curved bridges to earthquake-induced pounding. The model is
based on a recent work of the authors conducted to study the rigid-body motion of horizontally symmetric
curved bridges subjected to earthquake-induced pounding. [1]. In this paper, this model has been employed to
perform a detailed parametric analysis to particularly study the influence of different parameters on the torsional
response of a prototype curved bridge affected by seismic pounding.

2. Mathematical Modeling of the Problem

The curved geometry of the deck of a horizontally curved bridge diverts the direction of the reaction forces of
the columns from the principal axes of the bridge during an earthquake excitation that couples the torsional
motion of the deck to its translational motions. This coupled motion may impose much worse condition on the
motion of the deck when it is associated with seismic pounding between the deck and the abutments. Fig. 1
shows the dynamic model used to deal with such features inherent in the dynamic behavior of horizontally
curved bridges [1]. This model is valid under these basic assumptions: (i) the deck is rigid [3,4,18,19]; (ii) the
soil-structure interaction is negligible [20]; (iii) the bearings used at the abutments act as ideal roller supports;
(iv) the mass of the bridge is entirely attributed to the mass center of the deck [21]; and (v) the shear-
displacement relations of the columns are bilinear but their torsional moment-rotation relations are linear.

2
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

2.1 Dynamic model

The general geometrical features of the proposed dynamic model are shown in Fig. 1. The in-plane rigid-body
motion of the deck subjected to earthquake-induced pounding is formulated in the xy-coordinate system.
However, the restoring forces of the columns and the impact forces between the deck and the abutments are
described in the r- and nt-coordinate systems, respectively, as shown in Fig. 1. The curved geometry of the
deck is idealized by a circular arc represented by subtended angle (0<< ), radius R and width W. It is
assumed that the curvature center of the deck is coincident with the origin of the xy- and r-coordinate systems
(i.e. point O(0,0)). The degrees of freedom of the model, u ox , u oy , and , are assigned to this point. The y-axis is
the axis of symmetry of the deck. The mass center of the deck, C m (0,y m ), is located on this axis. The polar
coordinate of i-th column is C si (r si , si ) in which angle si is bounded in the interval | si /2|</2 [1]. The
restoring forces of i-th column is modeled with two bilinear translational springs with radial and azimuthal initial
stiffnesses k sri and k si , post- to pre-yield stiffness ratios sri and si , and yield displacements d ysri and d ysi ,
receptively, and a linear rotational spring with stiffness k si .

Fig. 1 3DOF dynamic model of typical horizontally curved bridges

The y-coordinate of the mass center, y m , and the radius of gyration of the deck about the curvature center,
r o , can be respectively calculated by the following expressions [1],
1
sin
1 2 2R 1 2 2
ym =
2+ , ro =
1 + (1)
6 2 4 2
where =L/W is aspect ratio of the deck (>1).

2.2 Restoring forces of columns

The restoring force vector of i-th column represented in the r-coordinate system is F si ={f sri , f si , t si }T. The
radial, f sri , and azimuthal, f si , shear forces and their interactions during the earthquake excitation are modeled by
the normalized form of Wang-Wen biaxial hysteresis model by disregarding its strength reduction effects [22].
However, for the sake of simplicity, the torsional moment, t si , is modeled by a linear spring. The effect of the
natural damping is disregarded; it will be taken into account later when the equation of motion of the bridge is
developed. Therefore, F si is given by [1],

3
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

=
Fsi (t) K si U si (t) + H si (t) (2)
where K si =Diag( sri k sri , si k si , k si ) is post-yielding stiffness matrix, U si (t)={u sri , u si , }T is deflection vector,
and H si (t)={h sri , h si , 0}T is hysteresis force vector in which h sri =(1- sri )f ysri z sri and h si =(1- si )f ysi z si where f ysri
and f ysi are radial and azimuthal yield forces; z sri and z si are dimensionless hysteresis variables given by the
following coupled nonlinear first-order differential equations [22],

A s u sri (| u sri || zsri |n 1 [s + s sgn(u sri z sri )]+ | u si || z si |n 1 [s + s sgn(u si z si )]s ) z sri
1
=z sri
d ysri
(3)
A s u si (| u si || zsi | [s + s sgn(u sri zsri )]s ) z si
1 n 1 n 1
=z si [s + s sgn(u si zsi )]+ | u sri || zsri |
d ysi

In Eq. (3), {A s , n, s , s } are parameters that control the shape and the size of the hysteretic loop, s
represents the biaxial interaction between the radial and azimuthal shear forces, and sgn(.) represents signum
function. These parameters are selected in such a way that they represent an ideal bilinear hysteresis model with
a biaxial coupled behavior along the r- and -axes [1]: A s =1, n=25, s = s =0.5, and s =1.

2.3 Impact forces

The normal impact forces between the deck and the abutments are modeled by nonlinear viscoelastic model,
known as Jankowski contact model [23,24]. This model consists of a nonlinear spring in parallel with a
nonlinear damper to simulate the kinetic energy absorption and dissipation, respectively. The damping part only
acts during the approach phase of collision, and consequently, most of the energy is dissipated in this phase. The
amount of energy dissipated during the restitution phase is negligible. Furthermore, the tangential impact force
(i.e. friction force) is modeled by Karnopp friction model [1,25], considering the effects of stick-slip
phenomenon at the contact nodes. It is assumed that the collisions between the deck and the abutments occur at
the corners of the deck denoted as #1, #2, #3, and #4 in Fig. 1. The contact forces at the left side for i-th (i=1 and
4) contact node are,
k pn pnli (t)3/ 2 + c pnli pnli (t) pnli (t) > 0 pnli (t) > 0

=
f pnli (t) k pn pnli (t)3/ 2 pnli (t) > 0 pnli (t) 0
0 pnli (t) 0

c pnli =
2pn Mk pn pnli (t) , pnli (t) = N Tpli U o (t) g , pnli (t) = (t)
N Tpli U o
(4)
k f pnli (t)sgn( ptli ) ptli vs0


f ptli (t) = f etli (t) ptli < vs0 s f pnli (t) > f etli , ptli (t) = (t)
TpliT U o

s f pnli (t)sgn(f etli ) ptli < vs0 s f pnli (t) f etli

in which N pli ={cos(/2), sin(/2), r pli }T, T pli ={sin(/2), +cos(/2), 0}T, r pl1 =R-W/2, and r pl4 =R+W/2. The
contact forces at the right side for i-th (i=2 and 3) contact node are,

4
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

k pn pnri (t)3/ 2 + c pnri pnri (t) pnri (t) > 0 pnri (t) > 0

=
f pnri (t) k pn pnri (t)3/ 2 pnri (t) > 0 pnri (t) 0
0 pnri (t) 0

c pnri =
2pn Mk pn pnri (t) , pnri (t) =
N Tpri U o (t) g , pnri (t) = (t)
N Tpri U o
(5)
k f pnri (t)sgn( ptri ) ptri vs0


f ptri (t) = f etri (t) ptri < vs0 s f pnri (t) > f etri , ptri (t) =T
Tpri U o (t)

s f pnri (t)sgn(f etri ) ptri < vs0 s f pnri (t) f etri

in which N pri ={cos(/2), +sin(/2), r pri }T and T pri ={+sin(/2), +cos(/2), 0}T, r pr2 =R-W/2, and r pr3 =R+W/2.
In Eqs. (4) and (5), pnai (t) and ptai (t) are relative normal and tangential displacements at the contact
nodes, pnai (t) and ptai (t) are relative normal and tangential velocities at the contact nodes (subscript a is
associated with the abutment number, i.e. on left side: {a1; i=1, 4} and on right side: {ar; i=2, 3}), g is the
size of the gap, k pn is impact stiffness parameter, and pn is damping ratio. In the friction model, s and k are
static and kinetic coefficients of friction, respectively, v s0 is a velocity limit with a small value below which the
velocity of the contact node is assumed to be zero when the contact node sticks, and f etai (t) is the external force
acting upon the deck at the contact node during the sticking phase, i.e. ptai (t) =
0 . The force will be formulated
in section 2.4.

2.4 Equation of motion

The equation of motion of the bridge is given by [1],

(t) + CU
MU (t) + KU (t) + H (t) + F (t) = (t)
MU (6)
o o o s p g

where=U o (t) {u o x , u o y , }T = (t) {u , u , }T , and=

, U (t) {u
U u oy ,
ox , }T are relative displacement, velocity
o ox oy o

and acceleration vectors of the curvature center of the deck, respectively, and U (t) = {u u ,0}T is ground
, g gx gy

acceleration vector. The rest of matrices and vectors used in this equation are defined as follows:
i) M is mass matrix of the bridge defined as,
M 0 My m
M = 0 M 0 (7)
My m 0 I o
where M is the total mass of the deck.
ii) K is post-yielding stiffness matrix of the bridge given as,
K XX K XY K rs sin s
=K K XY K YY + K rs cos s (8)
K rs sin s + K rs cos s K + rs2 K
where K XX , K XY , K YY , K , and K are defined as [1],

5
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

ns
=
K XX
i =1
sri k sri cos 2 si + si k si sin 2 si
ns
K=
XY (
i =1
sri k sri si k si )sin si cos si
ns
=
K YY
i =1
sri k sri sin 2 si + si k si cos 2 si (9)
ns
K=

i =1
si k si
ns
K =
i =1
si k si (rsi2 rs2 ) + k si

and (r s , s ) represents the location of the post-yielding stiffness center of the bridge in the r-coordinate system
defined as [1],
2 2
ns ns ns
si si si
k r sin si + si k si rsi cos si si k si rsi cos si
= i 1= i1
rs = , s cos 1 i =1 (10)
K K rs


where | s /2| < /2. The static radial stiffness eccentricity is defined as e sr0 =r s0 y m in which r s0 is the radial
coordinate of the pre-yield or static stiffness center (r s0 =r s in Eq. (10) when sri = si =1); and, the static azimuthal
stiffness eccentricity is defined as e s0 = s0 /2 in which s0 is azimuthal coordinate of pre-yield or static
stiffness center ( s0 = s in Eq. (19) when sri = si =1).
iii) C is damping matrix of the bridge defined based on Rayleigh damping method as follows,
2
=C ( 13M + K 0 ) (11)
1 + 3
in which K 0 is pre-yield stiffness matrix (K 0 =K when sri = sti =1 and i=1,2, , n s ), 1 and 3 are natural
circular frequencies of the first and third modes computed on the basis of K 0 , and is critical damping
coefficient assumed to be equal to 2.5%.
iv) H s (t)={h sx , h sy , s }T is total hysteresis force vector of columns in which vector components h sx , h sy , and s
are defined as follows,
ns
=h sx h
i =1
sri cos si h si sin si
ns
=
h sy h
i =1
sri sin si + h si cos si (12)
ns
s = h si rsi i =1

T
v) F p (t)={f px , f py , t p } is total impact force vector in the xy-coordinate system defined as,

P
=i 1,4
pli pnli
=i 2,3
pli ptli =
F (t) N f (t) + T f (t) N f
pri pnri (t) Tpri f ptri (t) (13)

The external force f etaj (t) applied to the deck when the j-th contact nodes on the left side (al; j=1, 4) or
the right side (ar; j=2, 3) sticks (i.e. when
ptaj (t) T
= Tpaj = 0 ) is given by [1],
U o (t)

6
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

( Tpaj M 1Tpaj ) Tpaj ( M 1CU o (t) + M 1KUo (t) + M 1Hs (t) + M 1FPSj (t) + U g (t) )
1
f etaj (t) = T T
(14)

where F PSj (t) is defined as [1],

PSj
=i 1,4
=F (t)
pli pnli
=i 2,3
pli ij ptli N f (t) + T (1 )f (t) N f
pri pnri (t) Tpri (1 ij )f ptri (t) (15)

in which ij is the Kronecker delta; ij =1 when i=j and ij =0 when ij.

3. Numerical Example
The proposed analytical model is utilized to investigate the seismic responses of a curved bridge prototype
subjected to earthquake-induced pounding. Fig. 2 shows geometrical details of this bridge. It is a prestressed
reinforced concrete curved bridge consisting of a box-girder deck and four single-column bents. The deck is
assumed to be rigid and monolithically connected to the bents. The main geometrical and dynamic parameters of
the bridge are [1,26]: =39.6, R=305 m, L=210.8 m, M=2894 tons, I o = 2693108 ton.m2, =20, r s =300.72 m,
y m =299 m, e sr0 =1.72 m, and e s0 =0 which implies that the bridge stiffness is symmetric with respect to the y-
axis, i.e. s =/2.

Fig. 2 Structural details of the curved bridge prototype; (a) plan, (b) longitudinal cross section, (c) transverse
cross-section, (d) columns cross-sections, (e) bilinear force-displacement model of columns [1,26].

7
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

Fig. 3 Absolute acceleration response spectra of seven pairs of ground motion records applied along the (a) x-
and (b) y-axes of the bridge model.

The properties of the bilinear force-displacement behaviors of the columns along the r- and -axes are
shown in Table 1. The dynamic model of the bridge is analyzed under a set of seven ground motion records
selected from the PEER Strong Ground Motion Database [27]. Fig. 3 shows the absolute acceleration response
spectrums of these ground motions and their geometric mean response spectra along the x- and y-axes. These
records have been scaled to a PGA=1.0g representing a very high-intensity ground shaking. The torsional
response of the model are taken as the mean of the responses calculated for each principal direction based on the
recommendation of AASHTO seismic design guideline [28]. The main parameters of the contact model used in
this study are the impact stiffness k pn , the damping ratio pn , the coefficients of static s and kinetic k friction,
and the limit velocity of Karnopp friction model v s0 . Table 2 presents the values of these parameters obtained
from different analytical and empirical methods developed for concrete-to-concrete frictional impacts [1]. The
equation of motion (i.e. Eq. (6)) is implemented in Matlab and is numerically solved by the 4th-order Runge-
Kutta method. In this paper, the size of the time-step is taken to be equal t=10-3 sec, which is not only
adequately small to avoid numerical instability and computational error, but large enough to increase the
computational speed [1].

Table 1 Parameters of the bilinear force-displacement behaviors of the columns along the r- and -axis [1].
Direction k s (kN/m) d ys (cm) s
Radial (r) 16000 8 0.18
Azimuthal () 72000 5 0.08

Table 2 Parameters of the contact model [1].

k pn (N/m1.5) pn s k v s0 (m/s)
2.75109 0.35 0.625 0.5 105

8
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

Fig. 4 Time histories of the torsional responses of the deck for the case of without pounding compared to those
of with pounding; (a) rotation, (b) angular velocity, and (c) angular acceleration of the deck.

In order to gain a general insight into the sensitivity of the torsional response of the deck to earthquake-
induced pounding, time histories of the rotation, the angular velocity, and the angular acceleration of the
curvature center of the deck for the case of pounding are compared to those for the case of without pounding.
Figs. 4(a-c) plot these results obtained from the nonlinear time-history analysis of the bridge under the x- and y-
components of 1995 Kobe earthquake while it has been assumed that g =5 cm, =39.6, and e sr0 =1.72 m. It is
observed that in contrast to the angular velocity of the deck, the rotation and the angular acceleration of the deck
are significantly affected by the seismic pounding. Fig. 4(a) shows that the absolute maximum rotation of the
deck is amplified by nearly 3 times from || max =0.04 for the case of no pounding to || max =0.11 for the case of
with pounding. The collision between the deck and the abutments results in a permanent clockwise rotation of
the deck as much as =0.037 whose magnitude is almost 40% larger than =+0.027 for the case of no
pounding. It can be seen from Fig. 4(b) that the angular velocity of the deck is not affected so much by the
seismic pounding. However, Fig. 4(c) shows that the angular acceleration of the deck is extremely sensitive to
the seismic pounding such that it experiences several pulses at the times of the impacts at the contact nodes. The
largest one has a magnitude as much as || max =17.5 deg/sec2 which is 20 times larger than the absolute maximum
angular acceleration for the case of no pounding || max =0.88 deg/sec2. Fig. 5(a-d) show time histories of the
normal impact forces at the contact nodes #1, #2, #3, and #4, respectively. As can be seen, the number of
impacts between the deck and the left abutment (i.e. impacts at contact nodes #1 and #4) during the earthquake is
greater than that for the right abutment (i.e. impacts at contact nodes #2 and #3). Furthermore, the magnitudes of
the normal impact forces for the left abutment are higher than those for the right abutment. These impact forces
produces a clockwise torsional moment about the curvature center which is the main cause of the permanent
clockwise rotation of the deck ( < 0), as shown in Fig. 4(a).

9
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

Fig. 5 Time histories of the normal impact forces at the contact nodes (a) #1, (b) #2, (c) #3, and (d) #4.

4. Parametric Study
The torsional response of horizontally curved bridges subjected to earthquake induced pounding is highly
dependent on the parameters of the impact forces, the geometry of the deck, and the stiffness distribution of the
columns. In order to identify the key parameters that affect these responses, a detailed parametric study is carried
out on the curved bridge prototype by varying different parameters, including the size of the gap between the
deck and the abutments, g , from 2.5 cm to 27.5 cm with a step of 2.5 cm, the subtended angle of the deck, ,
from 5 to 175 with a step of 17 (note that L is kept constant, L=R=210.8 m, implying that the curved bridges
considered with different are equivalent [28]), and the normalized static radial stiffness eccentricity ratio,
sr =e sr0 /r o , which can be simplified into following form,
ns

2 sin si
12 2 + 2 sin 2
=sr i =1
(16)
4 2 + 2
ns 3 4 +
2 2

It should be noted that the stiffness eccentricity defined here is based on its classical definition, i.e. it is measured
from the mass center of the deck. Eq. (16) can be manipulated to vary the location of the stiffness center relative
to the mass center which results in three cases: (1) sr <0 (asymmetric model, r s <y m ), (2) sr =0 (symmetric
model, r s =y m ), and (3) sr >0 (asymmetric model, r s >y m ). The horizontally curved bridge prototype is analyzed
in these three cases in order to investigate the influence of the stiffness eccentricity on the torsional response of
the deck. Two extreme arrangements are chosen for the columns on the plane of the bridge in the asymmetric
cases of 1 and 3, as shown in Fig. 6 (a) and (c); Fig. 6(b) also shows the arrangement of the columns in the
symmetric case, i.e. case 2. Fig. 6(d) shows the variation of sr with for these three cases. It is seen that sr
increases with in cases 1 and 3, but it is zero in case 2 ( sr =0). In order to keep this quantity equal to zero (i.e.
to match the location of the stiffness center with that of the mass center) in all values of considered, the
azimuthal coordinates of the columns, si (i=1, 2, 3, and 4), have to be varied by according to what shown in
Fig. 6(e). The torsional response quantities of interest are the absolute maximums of the rotation and angular
acceleration of the deck, i.e. max (||) and max (||), respectively. These response quantities are calculated for
each pair of ground acceleration records in Fig. 3 by varying parameters discussed above, and then averaged to
calculate the mean response quantities.

10
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

Fig. 6 (a-c) Sketches of cases 1, 2, and 3 considered for the parametric analysis (d) variation of the normalized
static radial stiffness eccentricity ratio, sr , with for cases 1, 2, and 3; (e) variation of the azimuthal coordinates
of the columns with for case 2.

Fig. 7 Absolute maximum torsional responses of the deck of the curved bridge prototype for different g and ;
(a-c) rotation of the deck for cases 1, 2, and 3, respectively; (d-f) angular acceleration of the deck for cases 1, 2,
and 3, respectively.
Figs. 7(a)-(c) show 2D image plots of the absolute maximum rotation of the deck versus g and for cases 1, 2,
and 3, respectively. It is seen that for g > 22.5 cm, beyond the dashed lines, the colors of the images do not
change with g implying that the seismic pounding does not occur when the size of the gap is larger than 22.5

11
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

cm. It is also clear that the rotation of the deck increases with sr from case 1 to case 3, i.e. when the distance
between the stiffness center and the curvature center is increased. The peak value of the rotation of the deck
occurs in case 3 and for and g approximately in intervals 5 < < 60 and 2.5 cm < g < 7.5 cm, respectively.
As the torsional moments of the columns are proportional to the rotation of the deck, t si =k si , we can expect
that the peak of these responses occur at the same values for and g . Figs. 7(d)-(e) show 2D image plots of the
absolute maximum angular acceleration of the deck versus g and for cases 1, 2, and 3, respectively. It is
observed that, in contrast to the rotation of the deck, the angular acceleration of the deck is not very sensitive to
the variation in the value of sr . The variation of this response quantity by g and has a quite similar trend in
the three cases. It can be generally concluded that the peak value of the angular acceleration of the deck in these
three cases also occurs for and g approximately in intervals 5 < < 90 and 2.5 cm < g < 7.5 cm,
respectively.
The results of the parametric analysis show that seismic pounding can significantly amplify the torsional
response of the decks of horizontally curved bridges. Therefore, there is a need to protect these bridges from
excessive in-plane rotation of the deck under earthquake-induced pounding. For example, seismic protective
devices can be useful for this purpose [17,29].

5. Conclusion
In this paper, the sensitivity of the torsional response of horizontally curved bridges subjected to earthquake-
induced pounding has been studied by a three-degree-of-freedom nonlinear model. In this model, the radial and
azimuthal shear forces of the columns and their bilateral interactions have been modeled by a coupled-biaxial
bilinear hysteresis model. The normal and tangential impact forces at the corners of the deck (i.e. contact nodes)
have been modeled by Jankowski contact model and the Karnopp friction model, respectively. The bridge model
has been employed to analyze the torsional responses of a one-way asymmetric curved bridge prototype ( s =/2)
to the deck-abutment collisions during strong ground motions. It has been shown that the rotation and the
angular acceleration of the deck are more sensitive to seismic pounding than the angular velocity of the deck.
Finally, a parametric analysis has been conducted on the curved bridge prototype by varying different parameters
of the model, including the size of the gap between the deck and the abutments g , the subtended angle of the
deck , and the normalized static radial stiffness eccentricity sr . It has been concluded that the rotation of the
deck, and consequently, the torsional moments of the columns increase with the increase in the distance between
the stiffness center and the curvature center of the deck. The peak value of these responses occur for the curved
bridges with 5 < < 60 and 2.5 cm < g < 7.5 cm when their stiffness centers are located above their mass
centers ( sr > 0).

6. References
[1] Amjadian M, Agrawal AK (2016): Rigid-body motion of horizontally curved bridges subjected to earthquake-induced
pounding. Bridge Engineering, 10.1061/(ASCE)BE.1943-5592.0000962, 04016090.
[2] Linzell D, Hall D, White D (2004): Historical perspective on horizontally curved I girder bridge design in the United
States. Bridge Engineering, 9(3), 218229.
[3] Kalantari A, Amjadian M (2010): An approximate method for dynamic analysis of skewed highway bridges with
continuous rigid deck. Engineering Structures, 32(9), 28502860.
[4] Kalantari A, Amjadian M (2010): An approximate method for dynamic analysis of highway skewed bridges with
continuous deck. 9th US National and 10th Canadian Conference on Earthquake Engineering 2010, Including Papers
from the 4th International Tsunami Symposium vol. 5, Toronto, Canada.
[5] Amjadian M, Kalantari A, Agrawal AK (2016): Analytical study of the coupled motions of decks in skew bridges with
the deck-abutment collision. Journal of Vibration and Control, 10.1177/1077546316659781.
[6] Kawashima K (2012): Damage of bridges due to the 2011 great east Japan earthquake. Proceedings of the International
Symposium on Engineering Lessons Learned from the 2011 Great East Japan Earthquake, Tokyo, Japan.

12
 16th World Conference on Earthquake, 16WCEE 2017
Santiago Chile, January 9th to 13th 2017

[7] Jennings PC (1971): Engineering features of the San Fernando earthquake of February 9. Technical Report California
EERL-71-02, California Institute of Technology Earthquake Engineering Research Laboratory, California, USA.
[8] Bruneau M (1998): Performance of steel bridges during the 1995 HyogokenNanbu (Kobe, Japan) earthquakea
North American perspective. Engineering Structures, 20(12), 10631078.
[9] Falamarz-Sheikhabadi MR, Zerva A, Ghafory-Ashtiany M (2016): Mean absolute input energy for in-plane vibrations
of multiple-support structures subjected to non-stationary horizontal and rocking components. Probabilistic
Engineering Mechanics, 45, 87101.
[10] Falamarz-Sheikhabadi MR, Ghafory-Ashtiany M (2015). Rotational components in structural loading. Soil Dynamics
and Earthquake Engineering, 75, 220233.
[11] Kawashima K, Takahashi Y, Ge H, Wu Z, Zhang J (2009): Reconnaissance report on damage of bridges in 2008
Wenchuan, China, earthquake. Earthquake Engineering, 13(7), 965996.
[12] Han Q, Du X, Liu J, Li Z, Li L, Zhao J (2009): Seismic damage of highway bridges during the 2008 Wenchuan
earthquake. Earthquake Engineering and Engineering Vibration, 8(2): 263273.
[13] Sun Z, Wang D, Guo X, Si B, Huo Y (2012): Lessons learned from the damaged Huilan Interchange in the 2008
Wenchuan earthquake. Bridge Engineering, 17(1), 1524.
[14] Williams D, Godden W (1979): Seismic response of long curved bridge structures: Experimental model studies.
Earthquake Engineering & Structural Dynamics, 7(2), 107128.
[15] Ijima K, Obiya H, Aramaki G, Kawasaki N (2001): A study on preventing the fall of skew and curved bridge decks by
using rubber bearings. Structural Engineering and Mechanics, 12(4), 347362.
[16] Wieser J, Zaghi A, Maragakis M, Buckle I (2012): A Methodology for the experimental evaluation of seismic pounding
at seat-type abutments of horizontally curved bridges. Structures Congress, Chicago, Illinois.
[17] Ruiz Julian FD, Hayashikawa T, Obata T (2007): Seismic performance of isolated curved steel viaducts equipped with
deck unseating prevention cable restrainers. Constructional Steel Research, 63(2), 237253.
[18] Kalantari A, Amjadian M (2011): Investigation of the seismic pounding effect on torsional response of highway
bridges. 6th International Conference of Seismology and Earthquake Engineering, Tehran, Iran.
[19] Amjadian M, Kalantari A (2012): Influence of seismic pounding on dynamic response of skewed highway bridges. 15th
World Conference on Earthquake Engineering, Lisbon, Portugal.
[20] Mylonakis G, Gazetas G. (2000): Seismic soil-structure interaction beneficial or detrimental? Earthquake Engineering,
4(3), 277301
[21] Meng JY, Lui EM, Liu Y (2001): Dynamic response of skew highway bridges. Earthquake Engineering, 5(2), 205
223.
[22] Wang CH, Wen YK (2000): Evaluation of Pre-Northridge low-rise steel buildings, I: Modeling. Structural
Engineering, 126(10), 11601168.
[23] Jankowski R (2005): Non-linear viscoelastic modelling of earthquake-induced structural pounding. Earthquake
Engineering & Structural Dynamics, 34(6), 595611.
[24] Jankowski R (2006): Analytical expression between the impact damping ratio and the coefficient of restitution in the
non-linear viscoelastic model of structural pounding. Earthquake Engineering & Structural Dynamics, 35(4), 517524.
[25] Karnopp D (1985): Computer simulation of stick-slip friction in mechanical dynamic systems. Dynamic Systems,
Measurement, and Control, 107(1), 100-103.
[26] Tondini N, Stojadinovic B (2012): Probabilistic seismic demand model for curved reinforced concrete bridges. Bulletin
of Earthquake Engineering, 10(5), 14551479.
[27] PEER (2013): PEER Ground Motion Database.
[28] AASHTO (2011): Guide Specifications for LRFD Seismic Bridge Design. American Association of State Highway and
Transportation Officials, Washington, USA.
[29] Agrawal AK, Amjadian M (2016): Seismic Component Devices, Innovative Bridge Design Handbook edited by
Pipinato A, 1st edition, Butterworth-Heinemann, Boston, 531553.

13